The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (e.g., python3 scripts/crypto.py ticker BTC/USDT) that incorporate user-supplied parameters like trading pairs, exchange names, and alert thresholds. While the skill includes guardrails to validate these inputs, the construction of shell commands from external input is a known risk for command injection if validation logic is circumvented.
[DATA_EXFILTRATION]: The skill accesses a local configuration file located at ~/.config/crypto/alerts.json to manage user alerts. Although it explicitly forbids network exfiltration in its guardrails, the combination of file access and the use of network-capable libraries (CCXT) constitutes a data exfiltration surface.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it processes untrusted data from both user inputs and external cryptocurrency exchange APIs.
Ingestion points: Trading pairs, exchange names, and numeric thresholds provided by users; market ticker, order book, and OHLCV data fetched from remote exchange APIs.
Boundary markers: The skill does not define explicit delimiters (e.g., XML tags or unique markers) to separate instructions from external data.
Capability inventory: Execution of subprocesses via python3 and read/write access to the local file system for configuration management.
Sanitization: The skill includes explicit instructional guardrails directing the agent to validate that exchange names and trading pairs match expected formats before command construction.

crypto-hardened